Stool-Test Screening Linked to 43% Lower Colorectal Cancer Death Risk

A new analysis of Sweden's Stockholm-Gotland screening program finds that people who actually completed a mailed fecal occult blood test had a 43% lower risk of dying from colorectal cancer, once researchers corrected for two kinds of statistical distortion.

Participation, Not the Invitation Alone, Explains the Larger Risk Reduction

The finding comes from a cohort study by Johannes Blom, Lennarth Nyström, and Håkan Jonsson, published online August 20, 2026, in JAMA Network Open. The researchers followed 376,511 people invited into the program for up to 14 years and recorded 1,668 colorectal cancer deaths.

After adjusting for contamination bias and nonadherence, the team found that simply being invited to screening was associated with a 26% lower risk of dying from colorectal cancer. Actively participating, meaning the person returned a stool sample, was associated with a 43% lower risk. That is a wide gap between two numbers describing the same program.

It is also a jump from what an earlier, unadjusted look at the same program had shown. A 2024 companion study by an overlapping author team compared people invited early to the program against those invited late or not at all, without the contamination and nonadherence corrections, and found a 14% lower mortality rate ratio. "A previous evaluation of the programme showed that those who received an invitation to screening had a 14 per cent lower risk of dying from the disease," said Blom, the corresponding author and a senior consultant at Karolinska Institutet's Södersjukhuset. "We have now been able to show that the mortality is significantly lower than that among those who actually take part in the screening."

The three figures are not competing claims about the same thing. The 14% and 26% numbers describe what happened to everyone who was invited, whether or not they screened. The 43% number describes what happened only to the people who sent in a sample. Comparing the last two shows how much of the program's benefit is currently being left on the table by the people who never mail the kit back.

A 376,511-Person Cohort and the Third Who Never Returned a Sample

The scale of the underlying data is part of why the researchers believe the adjusted estimates. The cohort spans more than 376,000 adults tracked for up to 14 years using Swedish national health registers, and it recorded 1,668 colorectal cancer deaths during that period. Screening in the region has run since 2008, with kits mailed every two years to residents currently aged 60 to 74.

That last card is the one Blom returned to when asked what the study means for individual readers. "Although screening is offered free of charge and the test is simple to carry out, around a third of people do not submit a sample," he said. "Our study highlights the importance of taking part in colorectal cancer screening, and shows that it can actually save lives."

Why Contamination and Nonadherence Change the Numbers

The gap between 14%, 26%, and 43% is not noise. It reflects what each estimate is actually measuring, and why raw, unadjusted comparisons of invited versus non-invited groups tend to understate a screening program's true effect.

Two separate problems pull an unadjusted estimate away from the truth. Contamination bias happens when people in the comparison group get screened anyway, outside the program being studied, most often because they eventually receive an invitation of their own or seek testing independently. That makes the "unscreened" group look less unscreened than it is, which shrinks the apparent benefit of the program. Nonadherence works differently: people invited to screening who never send back a sample are still counted in the "invited" group in a simple analysis, even though they received none of the direct benefit of a completed test. That also dilutes the invited group's numbers, this time by mixing screened and unscreened outcomes together under one label.

Blom and colleagues used statistical methods designed to strip out both effects, which is why the paper is titled around contamination bias and nonadherence rather than around the topline percentage. The result separates two different questions that a public-health program has to answer with two different numbers: how much good does an invitation, sent to everyone, do on average, and how much good does the test itself do for the person who actually uses it.

The Karolinska Institutet's own summary of the findings is candid about the limits of that approach. The cohort is large and the follow-up is long, but the 26% and 43% figures come from statistical adjustment applied to observational data, not from a randomized trial, so some uncertainty about their exact size remains. The study was funded by the Swedish Cancer Society, the Swedish Research Council, and Region Stockholm, and its authors report no conflicts of interest.

What the analysis does establish clearly is where the program's remaining opportunity sits. It is not in the invitation process, which already reaches the population it is meant to reach. It is in the roughly one in three invited people who set the kit aside.